| 一类有效的变尺度分层BP学习算法 |
| |
| 引用本文： | 王正欧 张雪峰. 一类有效的变尺度分层BP学习算法[J]. 天津大学学报(自然科学与工程技术版), 1996, 29(3): 364-369 |
| |
| 作者姓名： | 王正欧 张雪峰 |
| |
| 作者单位： | 天津大学系统工程研究所 |
| |
| 摘 要： | 提出一类有效的改进ＢＰ算法，它采用分层的变尺度学习算法克服一般变尺度学习算法处理的矩阵线数过高、难于实用的问题；同时对势态项系数采用一种自适应的调整方法，避免因其选择不当影响学习收敛速度问题．计算实例表明算法的优越性．
|
| 关 键 词： | 神经网络 BP算法 变尺度法 学习算法 |
A CLASS OF EFFECTIVE HIERARCHICAL BP LEARNING ALGORITHMS BASED ON VARIABLE SCALE METHOD |
| |
| Wang Zheng'' ou. A CLASS OF EFFECTIVE HIERARCHICAL BP LEARNING ALGORITHMS BASED ON VARIABLE SCALE METHOD[J]. Journal of Tianjin University(Science and Technology), 1996, 29(3): 364-369 |
| |
| Authors: | Wang Zheng'' ou |
| |
| Abstract: | This paper presents a class of effective improved BP algorithms. The algorithms use hierarchical learning approach to overcome the problem of oversize dimensionality existing in the general learning algorithm based on variable scale method, and use a self-adjusting method to determine the momentum factor, which avoids the problem of affecting the convergent velocity of learning due to incorrect selection of the factor. Simulations show the superiority of the proposed algorithms. |
| |
| Keywords: | neural networks BP algorithm variable scale method |
| 本文献已被 CNKI 维普 等数据库收录！ |
